Hassabis's prediction is quite aggressive, but the biggest risk is that society's readiness can't keep up with the technology.

Hassabis predicts that the development speed of AGI will exceed expectations, potentially achieving around 2029 to 2030. He believes that current AI is a prelude to more powerful intelligence in the future, and as multimodal and autonomous decision-making capabilities mature, a key breakthrough in AGI within three years is no longer science fiction; but he warns that the global society is insufficiently prepared and needs to establish rules and safeguards in advance.